What file transfer speeds over 802.11g network?

I've got two laptops and a router all running 802.11g (g mode only, not
mixed). There are no other network devices visible anywhere in the
neighbourhood. I am copying some large .avi files between the Windows XP
laptops. Using Windows Task Manager I am monitoring transfer rates and am
amazed and disappointed to see that the network is only running at a max of
20% capacity and sometimes single digit %s. That's anything between about
300KB and 1.2MB per second. I've got no other apps running. Signal strength
is "Excellent" on both laptops and all three devices are in the same room.

Why is the throughput so miserable? Surely the disks can supply/receive data
quicker than that. I will try a direct ethernet connection when I get time
to see how that performs but surely I should be seeing somewhere near 100%
throughput rather than 3%-20%.

I also tried to set up an adhoc wireless network between the two laptops,
powering the router down altogether but despite following instructions in
Windows Help and Support I could not make any kind of connection between
them. Even though I tried to define manual connections and remove all
security they just wouldn't talk. Any obvious thoughts on this little
conundrum?

TIA,
Tim

--
Email address is munged. Please reply to newsgroup.

On Sun, 29 Feb 2004 18:27:28 -0000, "Tiny Tim" <_(E-Mail Removed)>
wrote:

>I've got two laptops and a router all running 802.11g (g mode only, not
>mixed). There are no other network devices visible anywhere in the
>neighbourhood. I am copying some large .avi files between the Windows XP
>laptops. Using Windows Task Manager I am monitoring transfer rates and am
>amazed and disappointed to see that the network is only running at a max of
>20% capacity and sometimes single digit %s. That's anything between about
>300KB and 1.2MB per second. I've got no other apps running. Signal strength
>is "Excellent" on both laptops and all three devices are in the same room.
>
>Why is the throughput so miserable?

Miserable throughput? You've got an average throughput.

You're seeing about 20Mbps data throughput (1.2MegaBytes per second *
8 bits to the Byte = 9.6 Megabits per second, which is being sent to
the Access Point by one PC, and then sent from the Access Point to the
other PC)

The maximum data throughput on a 802.11g network appears to be around
27 Megabits per second. This assumes that all the devices are
receiving a signal that's high enough quality (a factor that's not
only effected by signal strength) to allow them to run at 54Mbps, and
that all the devices on the network can hear each other.

Overheads, such as delays between data packets, packet preambles and
headers (parts of the 802.11 protocol which aren't visible to the end
user) account for the "missing bandwidth". The XP Task Manager
doesn't account for these, and, IMO, is somewhat misleading.

Derek
--
"All the world is an analog stage and digital circuits play bit parts."

Derek wrote:
> On Sun, 29 Feb 2004 18:27:28 -0000, "Tiny Tim" <_(E-Mail Removed)>
> wrote:
>
<snip>
> Miserable throughput? You've got an average throughput.
>
> You're seeing about 20Mbps data throughput (1.2MegaBytes per second *
> 8 bits to the Byte = 9.6 Megabits per second, which is being sent to
> the Access Point by one PC, and then sent from the Access Point to the
> other PC)
>
> The maximum data throughput on a 802.11g network appears to be around
> 27 Megabits per second. This assumes that all the devices are
> receiving a signal that's high enough quality (a factor that's not
> only effected by signal strength) to allow them to run at 54Mbps, and
> that all the devices on the network can hear each other.
>
> Overheads, such as delays between data packets, packet preambles and
> headers (parts of the 802.11 protocol which aren't visible to the end
> user) account for the "missing bandwidth". The XP Task Manager
> doesn't account for these, and, IMO, is somewhat misleading.
>
> Derek

Derek, thanks for the reply. It hadn't even occurred to me that the
theoretical 54Mbps would be shared between the two laptops but now I think
about it I guess that does seem obvious, if everything is running on the
same channel, which I suppose it must be.

I was also just doing a crude "divide bits by 10" to get bytes, allowing for
overheads. i.e. 54Mbps ~= 5.4 MBps. Perhaps that's not enough. Even so, that
means with shared bandwidth that the maximum per laptop,including overheads,
is going to be 2.7MBps. It just seems to me that there is a large gap
between 2.7MBps and the 1.2MBps that I am seeing as my maximum throughput
per laptop.

The laptops and AP are only about 6-8 feet apart with direct line of sight.
I find it hard to believe that overheads suck up over 50% of the total
bits/bytes transferred. Assuming I have relatively clean airspace, from an
RF point of view I still struggle to accept this level of performance as the
best I can hope for. I was hoping for more. I may try hard-wiring one of the
laptops to see if my throughput doubles with only one wireless device
attached.

Cheers,
Tim

--
Email address is munged. Please reply to newsgroup.

On Mon, 1 Mar 2004 07:55:39 -0000, "Tiny Tim" <_(E-Mail Removed)>
wrote:

>Derek wrote:
>> On Sun, 29 Feb 2004 18:27:28 -0000, "Tiny Tim" <_(E-Mail Removed)>
>> wrote:
>>
><snip>
>> Miserable throughput? You've got an average throughput.
>>
>> You're seeing about 20Mbps data throughput (1.2MegaBytes per second *
>> 8 bits to the Byte = 9.6 Megabits per second, which is being sent to
>> the Access Point by one PC, and then sent from the Access Point to the
>> other PC)
>>
>> The maximum data throughput on a 802.11g network appears to be around
>> 27 Megabits per second. This assumes that all the devices are
>> receiving a signal that's high enough quality (a factor that's not
>> only effected by signal strength) to allow them to run at 54Mbps, and
>> that all the devices on the network can hear each other.
>>
>> Overheads, such as delays between data packets, packet preambles and
>> headers (parts of the 802.11 protocol which aren't visible to the end
>> user) account for the "missing bandwidth". The XP Task Manager
>> doesn't account for these, and, IMO, is somewhat misleading.
>>
>> Derek
>

>I was also just doing a crude "divide bits by 10" to get bytes, allowing for
>overheads. i.e. 54Mbps ~= 5.4 MBps. Perhaps that's not enough. Even so, that
>means with shared bandwidth that the maximum per laptop,including overheads,
>is going to be 2.7MBps.

I wish it was that simple ... it's complicated by the fact that there
are "random" timeouts used to allow multiple devices a fair allocation
of bandwidth.

>The laptops and AP are only about 6-8 feet apart with direct line of sight.
>I find it hard to believe that overheads suck up over 50% of the total
>bits/bytes transferred.

Unfortunately, that is the case.

Proxim have a useful (although somewhat terse) paper on the factors
effecting throughput at:

http://www.proxim.com/learn/library/...erformance.pdf

> Assuming I have relatively clean airspace, from an
>RF point of view I still struggle to accept this level of performance as the
>best I can hope for. I was hoping for more. I may try hard-wiring one of the
>laptops to see if my throughput doubles with only one wireless device
>attached.

Practically, you might check if your cards and access point/router
allows you to switch to a 802.11g-only mode of operation - some
equipment seems to suffer noticably if "802.11b compatability" is
turned on.

Derek

--
'Listen,I'm holding my watch close to the phone so you can hear the
ticking. Fifty-seven, fifty-eight, fifty-nine - there, another gullible
idiot's just been born, you have company.'
Tom holt, my hero

Derek wrote:
> On Mon, 1 Mar 2004 07:55:39 -0000, "Tiny Tim" <_(E-Mail Removed)>
> wrote:
>
>> Derek wrote:
>>> On Sun, 29 Feb 2004 18:27:28 -0000, "Tiny Tim"
>>> <_(E-Mail Removed)> wrote:
>>>
>> <snip>
<snip>
> Practically, you might check if your cards and access point/router
> allows you to switch to a 802.11g-only mode of operation - some
> equipment seems to suffer noticably if "802.11b compatability" is
> turned on.
>
> Derek

I tried switching to g mode only and it appeared to make a small improvement
but nothing like the performance I was expecting.

I also had a browse of the Proxim article and it makes very disappointing
reading. The bottom line seems to be that the advertised 54Mbps performance
is barely more than marketing hype. What's worse is that 802.11g devices can
be slowed to "snails' pace" by the presence of 11b devices, thus making the
investment in 11g kit a complete waste of money.

I've got two 11g laptops and an 11b laptop (which is mostly powered off) but
it's annoying to discover that if the 11b laptop is fired up then the rest
of the network will suffer. I bought the 11b card to save some money, not
realising at the time it would make the premium paid for the 11g card a
relative waste of money. Ho hum :-(

--
Email address is munged. Please reply to newsgroup.

On Mon, 1 Mar 2004 10:53:33 -0000, "Tiny Tim" <_(E-Mail Removed)>
wrote:

<snip>

>I also had a browse of the Proxim article and it makes very disappointing
>reading. The bottom line seems to be that the advertised 54Mbps performance
>is barely more than marketing hype.

(As an aside I'd suggest that you don't look at the throughput on
Ethernet, which uses similar mechanisms to manage access to a shared
channel).

> What's worse is that 802.11g devices can
>be slowed to "snails' pace" by the presence of 11b devices, thus making the
>investment in 11g kit a complete waste of money.

This kind of problem is why a number of the professional access points
(as opposed to the consumer equipment that I'm assuming that we're
talking about here) use two radios, operating on different, "well
separated" channels, one for -g access and the other for -b access.

Derek
--
"English doesn't borrow from other languages. It follows them down dark
alleys and mugs them for grammar."